byted-livesaas-master

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.80). The skill's required workflow (see SKILL.md §11 "大模型联网检索（关键兜底）" and §9.7 doc-version retry) mandates using a web search and fetching/parsing public documentation pages (e.g., https://www.volcengine.com/docs and other web results) to determine API versions and drive retries, so the agent will ingest and act on content fetched from open/public web pages which can materially influence subsequent tool calls.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). The skill's onboarding explicitly runs network installs such as "npm i -g bytedlive-cli" (fetching packages from the npm registry) and may install via git+ssh from the internal byted-live-cli repo, which are runtime fetch-and-execute operations of a required CLI that the agent then runs to perform actions.